INDEX

Explanations

phrases related to actions or activities being performed by someone

oai_token-act-pair · gpt-3.5-turbo

instances of the word "went."

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

oint

-0.59

 affirmation

-0.58

oret

-0.56

ULTS

-0.56

opia

-0.56

ible

-0.55

ract

-0.55

 Camel

-0.55

 Monetary

-0.53

 representations

-0.53

POSITIVE LOGITS

 went

3.10

 goes

2.12

went

1.84

 flew

1.77

 took

1.71

 proceeded

1.67

 blew

1.66

 came

1.64

ran

1.61

 stayed

1.61

Activations Density 0.025%

phrases related to actions or activities being performed by someone

instances of the word "went."

No Comments

No Known Activations